SpringOne 2024 Top Secret Demo

By Josh Long, Cora Iberkleid, and Dr. Mark Pollack

Hi, Spring fans! Let's build something awesome for SpringOne 2024.

to run

• start docker containers and load db data (use init_db.sh)
• start the auth-service
• start the processor
• start the gateway
• start the service
• start the ui (use run.sh)

to test

• access the gateway at http://127.0.0.1:1010, this will redirect you to the UI for login
• log in using josh/secret or cora/secret - shhh... don't share the password!
• click to adopt the dog of your choice - beware the neurotic Prancer!

If you comment out the oauth2-resource-server in service/pom.xml, then you can also test the service app directly using:

curl -d'{"name":"Cora"}' -XPOST -H"content-type: application/json" http://localhost:8080/dogs/45/adoptions

architecture

Spring's Pooch Palace Dog Adoption Agency Architecture

• gateway - the first port of call for requests intended for our system
• ui - HTML and JavaScript application
• service - supports finding and adopting dogs.
• processor - a legacy system that is involved in the adoption process.
• auth-service - a Spring Authorization Server instance

the diagram is here

this has been a whirlwind tour of all things Spring

To recap, we've covered:

• Spring Boot 3.3 - the runtime sensation that's sweeping all nations
• Spring Data JDBC - you've got data, and we've got you covered
• Spring Modulith - we heard you like modules so we put some modularity in your modules
• Spring AI (new and improved, now with RAG support!) - to build the assistant
• Spring Integration - making friends since 2007
• Spring Authorization Service - your one-stop-shop for all things authentication and authorization
• Spring Gateway - handle cross-cutting network concerns with ease, aplomb, and alacrity

Beats

• We wanted a theme for our keynote but didnt want to rehash the tired old Spring Petclinic example. but we do like pets
• Let’s build a pet adoption agency
• Who doesn’t like dogs?
• I love dogs. I have a dog. Hes terribad. Reminds me of Prancer.
• All dogs, even the bad ones, deserve a good home
• start.spring.io GraalVM, Web, Data JDBC, PGVector, OpenAI, Modulith, AMQP, and OAuth Resource Server, etc.
• go to build and comment out the Resource Server dependency
• specify SQL DB connectivity credentials
• Add a package for adoptions
  • Build a controller that lets you post to update the owner field of the Dog to reflect the new owner
• We finished this but there's a lot of work involved in adopting a dog. We need to update other modules.
• Let’s build a kennel service.
• demo ApplicationModulelistener
• We can enable republished events and initialize the schema. Watch. Now we publish the event and then pull the plug on the machine and the event is republished
• Nice! We have a well structured system.
  • We can ensure it stays that way using ApplicationModules.verify()., Aldo get living documentation with Documenter(ApplicationModules)
• We publish the message to another module in the same codebase. But what if the kennel service was actually a different microservices? It’s an older Spring Boot module using 2.7 and Spring Integration.
• Easy. Add @Externalized to the message ad then add the requisite amqp or kafka starter
  • Make sure Externalized#target points to the destination in the broker
  • Nice! The other service got the message. Although, looking at it, it’s using the creakingly ancient Java 17 and Spring Boot 2.7. Something will have to be done about that…
  • OH WELL. for another time, surely.
• Anyway! We were able to adopt prancer. But we knew his ID. It’s 45. How is someone else to find him?
• We need an assistant to help find him
• Spring AI to the rescue
• Dr. Pollack enters the chat and gives us a brief overview of the history of AI (and scares off a smelly Python developer or something lol)
• Define the ChatClient
• Give it a user prompt. AI responds "im an AI, we dont have pets."
• Give it a system prompt about being an assistant at Springs Pet emporium . AI responds with a disappointed message, but at least it sounds like it should be able to do this job.
• VectorStores
• introduce the RAG pipeline (QuestionAnswerAdvisor)
• Nice! It found Prancer, but how is the rest of the program to work with this news?
• introduce output mapping and schema. (DogAdoptionSuggestion.class)
• I would also love it if the LLM could schedule a pick up time (the new function support gets used here)